U.S. patent application number 13/729105 was filed with the patent office on 2013-08-15 for system and method for improved forensic analysis.
This patent application is currently assigned to ChemImage Corporation. The applicant listed for this patent is ChemImage Corporation. Invention is credited to Arjun Bangalore, Jeffrey Beckstead.
Application Number | 20130208985 13/729105 |
Document ID | / |
Family ID | 43534869 |
Filed Date | 2013-08-15 |
United States Patent
Application |
20130208985 |
Kind Code |
A1 |
Beckstead; Jeffrey ; et
al. |
August 15, 2013 |
System and Method for Improved Forensic Analysis
Abstract
The present disclosure provides for a system and method for
analyzing questioned documents. A sample document is illuminated to
thereby generate a first plurality of interacted photons. The first
plurality of interacted photons are detected at a first detector to
thereby generate a digital image. The digital image is analyzed to
thereby identify at least one region of interest of the sample
document. This region of interest is illuminated to thereby
generate a second plurality of interacted photons. This second
plurality of interacted photons are passed through a tunable filter
and detected at a second detector to thereby generate a
hyperspectral image representative of the region of interest. The
hyperspectral image may then be analyzed to evaluate changes to or
differentiate different inks present in the sample document.
Chemometric techniques such as k-means clustering, PCA, and/Or
PLSDA may also be applied.
Inventors: |
Beckstead; Jeffrey;
(Valencia, PA) ; Bangalore; Arjun; (Monroeville,
PA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
ChemImage Corporation; |
|
|
US |
|
|
Assignee: |
ChemImage Corporation
Pittsburgh
PA
|
Family ID: |
43534869 |
Appl. No.: |
13/729105 |
Filed: |
December 28, 2012 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12806039 |
Aug 4, 2010 |
|
|
|
13729105 |
|
|
|
|
Current U.S.
Class: |
382/191 |
Current CPC
Class: |
G01N 2021/6423 20130101;
G06K 9/00442 20130101; G01N 21/55 20130101; G01J 3/02 20130101;
G01N 21/27 20130101; G01J 3/32 20130101; G06K 9/4652 20130101; G01J
3/0248 20130101; G01N 21/6456 20130101; G01J 3/28 20130101; G01J
3/10 20130101; G01N 21/4738 20130101; G01N 2021/6493 20130101 |
Class at
Publication: |
382/191 |
International
Class: |
G06K 9/00 20060101
G06K009/00 |
Claims
1. A method comprising: (a) selecting a first location of a sample
document, wherein the first location comprises at least one ink
type; (b) generating at least one hyperspectral image
representative of the first location; (c) analyzing the
hyperspectral image to determine whether or not two or more ink
types are distinguishable in the first location.
2. The method of claim 1 wherein in the analyzing further
comprises: (d) if two or more inks are distinguishable in the first
location, reporting a result, and (e) if two or more inks are not
distinguishable in the first location, (i) applying a first chemo
metric technique to generate a processed image, and (ii) analyzing
the processed image to determine if two or more inks are
distinguishable in the first location.
3. The method of claim further comprising, if two or more inks are
not distinguishable in the first location, repeating step (e) for
at least one other chemometric technique.
4. The method of claim 2 wherein the first chemometric technique
further comprises at least one of: principle component analysis,
partial least squares discriminate analysis, cosine correlation
analysis, Euclidian distance analysis, k-means clustering,
multivariate curve resolution, band t. entropy method, mahalanohis
distance, adaptive subspace detector, spectral mixture resolution,
and combinations thereof.
5. The method of claim 3 wherein the other chemometric technique
further comprises at least one of: principle component analysis,
partial least squares discriminate analysis, cosine correlation
analysis, Euclidian distance analysis, k-means clustering,
multivariate curve resolution, band t, entropy method, mahalanobis
distance, adaptive subspace detector, spectral mixture resolution,
and combinations thereof.
6. The method of claim 1 wherein analyzing the hyperspectral image
further comprises automatically detecting at least one of: ink
pixels, background pixels, and combinations thereof.
7. The method of claim 6 wherein the first chemometric technique is
applied only to ink pixels.
8. The method of claim 1 wherein the analyzing is at least
semi-automated.
9. The method of claim 1 wherein selecting the first location of
the sample document further comprises: generating a digital image
representative of the sample document, and analyzing the digital
image to select the first location.
10. The method of claim 1 wherein the hyperspectral image further
comprises at least one of: a visible hyperspectral image, a
fluorescence hyperspectral image, an infrared hyperspectral image,
a visible-near infrared hyperspectral image, and combinations
thereof.
11. The method of claim 1 wherein the analyzing further comprises
visual inspection by a user.
12. The method of claim 1 further comprising extracting spectral
information from the hyperspectral image and evaluating at least
one spectra associated with at least one ink present in the first
location.
13. The method of claim further comprising repeating steps (a)-(c)
for at least one other location in the sample document.
14. The method of claim 2 further comprising repeating steps
(d)-(e) for at least one other location in the sample document.
15. The method of claim 1 wherein generating the hyperspectral
image further comprises: illuminating the first location to
generate at least one plurality of interacted photons; passing the
interacted photons through a tunable filter to filter the
interacted photons into a plurality of wavelength bands; and
detecting the filtered interacted photons to generate the
hyperspectral image.
16. The method of claim 15 wherein the illuminating further
comprises the use of at least one of: a UV light source, a
broadband light source, and combinations thereof.
17. The method of claim 3 wherein the first chemometric technique
and the at least one other chemometric technique are different.
18. The method of claim 3 wherein the first chemometric technique
and the at least one other chemometric technique are the same.
19. The method of claim 13 wherein the first location of the sample
document and the at least one other location of the sample document
at least partially overlap.
20. The method of claim 13 wherein the first location of the sample
document and the at least one other location of the sample document
do not overlap.
Description
RELATED APPLICATIONS
[0001] This application is a divisional application of pending U.S.
patent application Ser. No. 12/806,039, entitled "System and Method
for Improving Forensic Analysis," which was filed on Aug. 4, 2010
and claims priority under 35 U.S.C. .sctn.119(e) to U.S.
Provisional Patent Application No. 61/231,077, filed on Aug. 4,
2009, entitled "Systems and Methods for Improved Forensic
Analysis." These applications are hereby incorporated by reference
in their entireties.
BACKGROUND
[0002] Forensic analysis involves the observation and
identification of an object that may exist in part or in its
entirety on some sort of supporting surface. This analysis
typically compares the sample in question to other possible
reference samples or reference data to make an association that
relates it to a specific person, object, place or event. Forensic
analysis is widely used in law enforcement or legal disputes as
evidence in a range of situations from homicide to fraud. More
specifically, the goal is usually to provide evidence of the
existence of a direct link, for example, between a suspect and a
crime scene, a victim and a suspect, a weapon and a suspect, etc.
To do so with a high degree of specificity and discrimination from
possible variations of the sample is essential. Examples of
forensic samples include, but are not limited to, fingerprints,
gunshot residues, condom lubricants, multi-layer paint chips,
fibers, ink samples and thin layer chromatography plates.
[0003] The quality of a forensic analysis is critical in making the
association of evidence as unambiguous as possible, thereby
providing compelling identifications and linkages. In many cases,
such as with fingerprints, this identification has widely accepted
requirements where as in others, such as fiber characterization and
comparison, the uniqueness of the results can be disputed. Even the
most unique and definitive identification of biological evidence
based on genetic information has been successfully questioned and
removed as compelling evidence. Minimizing the subjective
components or features of a forensic analysis to make compelling
identifications and linkages therefore becomes a critical aspect of
all forensic analysis. Doing so quickly and in a cost effective
manner is equally important.
[0004] In most legal cases, the ability of a jury or judge to
understand the forensic evidence, and the ability of the scientist
to convey its value determines the utility of the forensic method.
As a result, methods which allow the objects to be visually
compared or which show simple representations of the item under
scrutiny are the most widely accepted and understood by
non-specialists. Despite the existence of many advanced scientific
techniques and analysis methods that are very sophisticated, many
such techniques may not be understood by non-specialists, and may
thereby raise some doubts as to its validity. Visual forensic
analysis and visual comparisons are amongst the most widely
accepted forensic methods used to date.
[0005] Advances in science and technology have enabled many new
approaches to sample analysis, bringing forensic science into an
era which goes far beyond the classic perception of an investigator
looking thru a magnifying glass for small traces of evidence.
Numerous techniques exist that allow detailed chemical and
elemental identification. This includes most all analytical
chemistry methods, such as mass spectroscopy, x-ray analysis,
scanning electron microscopy and chromatography that are widely
used today to characterize gaseous, liquid and solid materials.
Many of these methods are extremely sensitive and require finite
material for their use that is consumed as part of the analysis
process. Advances in the sensitivity of analytical chemistry
methods and instruments over the years have reduced this problem
but these methods are still not considered non-destructive. This
becomes increasingly important as smaller and smaller pieces of
pieces of evidence are examined and required in forensic analysis.
The examination of questioned documents consists of the analysis
and comparison of questioned handwriting, hand printing, type
writing, commercial printing, photocopies, papers, inks, and other
documentary evidence with known material in order to establish the
authenticity of the contested material as well as the detection of
alterations. Forensic document examiners (FDEs) conduct
examinations on a wide variety of evidence including medical
records, wills, accounting records, checks, money and security
documentation such as passports and visas.
[0006] Document examination may include: detecting forgeries,
alterations, obliterations, additions or insertions, deletions, or
other changes to documents; the identification or elimination of
writing and/or printing with a known individual's writing;
restoration of burned, faded, or water soaked documents; the
identification, elimination, or classification of typewriting,
typewriters, and elements, ribbons and correction materials; paper
and ink analysis; dating of documents; searching documents for
impressed wiring, typing, or other identifying impressions;
photocopier identification, classification or comparison; rubber
stamps or their impressions; anything to do with the production of
documents or authenticity; among others.
[0007] Forensic document examiner conducts examinations on a wide
variety of evidence. This can range from examining medical records
to determine whether the entries were made at the time alleged,
examining anonymous letters being sent by one employee to another,
determined whether the signature on the will is the decedent's,
deciphering an entry that has been covered with white-out,
determining whether a particular photocopy was produced on a
suspected machine and so on.
[0008] The examination of questioned documents consists of the
analysis and comparison of questioned handwriting, hand printing,
type writing, commercial printing, photocopies, papers, inks, and
other documentary evidence with known material in order to
establish the authenticity of the contested material as well as the
detection of alterations.
[0009] Other techniques may include: thin layer chromatography TLC)
and optical spectroscopy. TLC is a type of liquid chromatography
that can separate chemical compounds of differing structure based
on the rate at which they move through a support under defined
conditions. TLC plates are used for separating color components of
printing and writing inks and indirectly for ink dating. The unique
way in which different inks separate into their individual
components allows an examiner to identify the formulation and
manufacturer. The manufacturer can then provide the earliest date
at which this formulation was sold, providing some dating
information. In addition to questioned document analysis, TLC
analysis can be used for the analysis of trace evidence, explosive
residue, insecticides, food toxins, biological materials and much
more.
[0010] Spectroscopic imaging combines digital imaging and molecular
spectroscopy techniques, which can include Raman scattering,
fluorescence, photoluminescence, ultraviolet, visible and infrared
(including SWIR, NIR, MWIR, and LWIR) absorption spectroscopies.
Raman spectra exhibit numerous features specific to molecular
structure and can provide valuable "fingerprints" for comparing and
differentiating materials. Surface enhanced resonance Raman
scattering (SERRS) may also hold potential. When applied to the
chemical analysis of materials, spectroscopic imaging is commonly
referred to as chemical imaging. Instruments for performing
spectroscopic (i.e. chemical) imaging typically comprise an
illumination source, image gathering optics, focal plane array
imaging detectors and imaging spectrometers.
[0011] In general, the sample size determines the choice of image
gathering optic. For example, a microscope is typically employed
for the analysis of sub micron to millimeter spatial dimension
samples. For larger objects, in the range of millimeter to meter
dimensions, macro lens optics are appropriate. For samples located
within relatively inaccessible environments, flexible fiberscope or
rigid borescopes can be employed. For very large scale objects,
such as planetary objects, telescopes are appropriate image
gathering optics.
[0012] For detection of images formed by the various optical
systems, two-dimensional, imaging focal plane array (FPA) detectors
are typically employed. The choice of FPA detector is governed by
the spectroscopic technique employed to characterize the sample of
interest. For example, silicon (Si) charge-coupled device (CCD)
detectors or CMOS detectors are typically employed with visible
wavelength fluorescence and Raman spectroscopic imaging systems,
while indium gallium arsenide (InGaAs) FPA detectors are typically
employed with near-infrared spectroscopic imaging systems.
[0013] Spectroscopic imaging of a sample can be implemented by one
of two methods. First, a point-source illumination can be provided
on the sample to measure the spectra at each point of the
illuminated area. Second, spectra can be collected over the an
entire area encompassing the sample simultaneously using an
electronically tunable optical imaging filter such as an
acousto-optic tunable filter (AOTF) or a liquid crystal tunable
filter ("LCTF"). Here, the organic material in such optical filters
are actively aligned by applied voltages to produce the desired
bandpass and transmission function. The spectra obtained for each
pixel of such an image thereby forms a complex data set referred to
as a hyperspectral image which contains the intensity values at
numerous wavelengths or the wavelength dependence of each pixel
element in this image.
SUMMARY
[0014] Visible to near infrared (NIR) reflectance/absorbance,
fluorescence imaging, optical microscopy and photography may be
used as tools by questioned document examiners to identify, capture
and characterize questioned documents. The present disclosure
provides for extensions of these systems and methods to
hyperspectral imaging ("HSI"). HSI may improve visualization and
provide additional information about the sample's formulation. HSI
combines standard digital imaging techniques with common
spectroscopic methods to provide increased sensitivity and
discrimination capabilities over traditional methods of questioned
document analysis. HSI is nondestructive and requires little to no
sample preparation, therefore decreasing the chances of possible
contamination.
[0015] The advantage of HSI lies in the information embedded within
the image. Because the images are a series of snapshots collected
as a function of wavelength, each pixel within the image has a
fully resolved spectrum associated with it. This method of data
collection allows for better visualization and discrimination of a
wider range of documents. Because the type of information contained
in hyperspectral images is easier to interpret for non-scientists,
the presentation of data and information to jurors in the courtroom
is easy and straightforward.
[0016] A hyperspectral image is obtained by collecting digital
images as a function of wavelength through the use of an
electro-optic tunable imaging filter. Images can be collect through
the visible to NIR range (400-1100 nm) in reflectance mode
(broadband white light illumination) or through the visible range
(400-720 nm) in fluorescence made (300-400 nm UV excitation).
Because the images are collected as a function of wavelength, each
pixel within the image has a fully resolved reflectance or
fluorescence spectrum associated with it. Following data
acquisition, image processing software may be used to generate
univariate (i.e. single wavelength) images characteristic of
absorbance, reflectance, or fluorescence exhibited by inks of
interest. This imaging software may comprise Chemlmage Xpert.RTM.
available from Chemlmage Corporation, Pittsburgh, Pa.
Alternatively, through the use of multivariate (i.e., multiple
wavelengths) statistical tools, the software takes advantage of the
viability present at a range of wavelengths to enhance image
contrast and reveal minor differences between inks. Conventional
imaging-based document analysis systems that rely on single
wavelength images are not capable of this superior level of image
enhancement.
[0017] HSI differs from the conventional method of document
examination in that it may utilize a multi-conjugate filter (MCF),
based on liquid crystal technology, in place of a general barrier
filter/camera configuration. The MCF is a computer controlled,
electro-optic device that can be tuned to transmit any discreet
wavelength across the visible or NIR region of the spectrum (i.e.,
400-1100 nm). Potential benefits of using an MCF are more fully
described in U.S. patent application Ser. No. 12/765,188, filed on
May 24, 2010, entitled "System and Method for Improved Forensic
Analysis", which is hereby incorporated by reference in its
entirety. Because different brands and types of inks have varying
formulations, the absorbance, reflectance or luminescence
properties of the inks will differ.
[0018] The present disclosure relates generally to forensic
analysis of questioned documents. More specifically, the present
disclosure provides for a system and method for analyzing
questioned documents using hyperspectral imaging. The system and
method provided for herein hold potential for distinguishing
between ink types present in a sample document. The ability to
distinguish between ink types may hold potential for determining
whether or not a sample document has been altered in some way.
[0019] The system and method of the present disclosure overcome
shortcomings present in the prior art. The system and method
provided for herein allow for quick and accurate analysis of sample
document based on visual inspection by a user. Being able to
visually distinguish between ink types present in a sample document
allows for results that can be easily interpreted.
BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The accompanying drawings, which are included to provide
further understanding of the disclosure and are incorporated in and
constitute a part of this specification, illustrate embodiments of
the disclosure and, together with the description, serve to explain
the principles of the disclosure.
[0021] FIG. 1 is a schematic representation of an exemplary
packaging option of a system of the present disclosure.
[0022] FIG. 2A is a schematic representation of a system of the
present disclosure.
[0023] FIG. 2B is a schematic representation of a various
illumination configurations of a system of the present
disclosure.
[0024] FIG. 3 is representative of a method of the present
disclosure.
[0025] FIG. 4 is representative of a method of the present
disclosure.
[0026] FIG. 5 is representative of a method of the present
disclosure.
[0027] FIG. 6A is illustrative of exemplary results of a method of
the present disclosure.
[0028] FIG. 6B is illustrative of exemplary results of a method of
the present disclosure.
[0029] FIG. 7A is illustrative of exemplary results of a method of
the present disclosure.
[0030] FIG. 7B is illustrative of exemplary results of a method of
the present disclosure.
[0031] FIG. 8A is illustrative of exemplary results of a method of
the present disclosure.
[0032] FIG. 8B is illustrative of exemplary results of a method of
the present disclosure.
[0033] FIG. 8C is illustrative of exemplary results of a method of
the present disclosure.
DETAILED DESCRIPTION
[0034] Reference will now be made in detail to the embodiments of
the present disclosure, examples of which are illustrated in the
accompanying drawings. Wherever possible, the same reference
numbers will be used throughout the drawings to refer to the same
or like parts.
[0035] The patent or application file contains at least one drawing
executed in color. Copies of this patent or patent application
publication with color drawing(s) will be provided by the Office
upon request and payment of the necessary fee
[0036] The present disclosure provides for a system for analyzing
questioned documents. FIG. 1 is illustrative of an exemplary
packaging option of a system of the present disclosure. FIG. 2A is
a schematic representation of a'system of the present disclosure.
The system 200 may be referred to commercially as the "HSI
Examiner" available from Chemlmage Corporation, Pittsburgh, Pa. The
system may be used for ink discrimination studies, to identify
alterations or forgeries, visualize hidden security features,
examine travel documents, and imaging TLC plates, among other
applications.
[0037] The system 200 comprises an illumination source (illustrated
in FIG. 2B). In one embodiment, the illumination source may
comprise a broadband white light illumination source. In one
embodiment, illustrated by FIG. 2B, the broadband white light
illumination may be configured to illuminate the sample document at
a plurality of different angles. In one embodiment, an illumination
source may be configured so as to illuminate a sample document or
region of interest with oblique illumination. In one embodiment,
the system 200 may also comprise at least one ultraviolet light
source. This ultraviolet light source may comprise at least one of:
UV-A, UV-B, UV-C, and combinations thereof.
[0038] Referring again to FIG. 2A, the system 200 comprises a
surface 210 for placing the sample document 215 for analysis. In
one embodiment, this surface may comprise a platform. In one
embodiment, the system may further comprise a light-tight
enclosure. The system 200 may also comprise collection optics for
collecting a plurality of interacted photons generated as a result
of illuminating the sample document 215. In FIG. 2A, the collection
optics is depicted as comprising a zoom lens 220. An embodiment
comprising a zoom lens may allow for varying focal length. In one
embodiment, the system 200 may be configured with collection optics
that allow for imaging approximately half a 8.5'.times.11' sheet of
paper with the ability to zoom up to 0.5.times.0.4 inches, with 600
dpi (121 p/mm) resolution. These configurations are provided as an
example of one embodiment of the present disclosure and do not
limit other embodiments of the system and method disclosed
herein.
[0039] The system 200 may also comprise a first detector,
illustrated in FIG. 2A as a RGB camera 225. In one embodiment, this
first detector may comprise a RGB color camera. RGB color cameras
hold potential for high resolution, court friendly digital imaging.
A means for directing the plurality of interacted photons through
said system 200 is illustrated in FIG. 2A as element 240. Element
240 may comprise a polarization beamsplitter and/or a mirror.
Element 240 may also be referred to as providing for "Imaging Mode
Selection" because it can direct the plurality of interacted
photons to either the first detector 225 for digital imaging or the
second detector 235 for hyperspectral imaging. In one embodiment,
the second detector 235 may comprise a spectral detector. In one
embodiment, the second detector 235 may be selected from the group
consisting of: a charge coupled device, a complementary
metal-oxide-semiconductor, a intensified charge coupled device, and
combinations thereof.
[0040] In one embodiment, discrete wavelengths of light may be
projected on to a charge-coupled device (CCD) camera. This may
provide up to 65 nm 535 channels of grayscale color, providing
significant sensitivity to any differences in ink or other features
of interest in a questioned document. In one embodiment, the second
detector may comprise a front-illuminated CCD with enhanced NIR QE
and/or a cooled CCD for reduced noise in long exposure fluorescence
experiments.
[0041] Hyperspectral imaging provides very high spectral
resolution, on the order of 5 nm, which allows the differentiation
of image components with subtle differences (i.e., black inks).
Combining high resolution spectral and spatial information allows
identification of these compounds on complex samples. In one
embodiment of the present disclosure, a spectral imaging range of
400 nm to 1100 nm may be used. In another embodiment, a spectral
tuning resolution of 1 nm may be used.
[0042] The system 200 may also comprise one or more tunable filters
230. In one embodiment, this filter may comprise a filter selected
from the group consisting of: a multi-conjugate tunable filter, a
liquid crystal tunable filter, acousto-optical tunable filters,
Lyot liquid crystal tunable filter, Evans Split-Element liquid
crystal tunable filter, Solc liquid crystal tunable filter,
Ferroelectric liquid crystal tunable filter, Fabry Perot liquid
crystal tunable filter, and combinations thereof.
[0043] In one embodiment, the hyperspectral imaging system of the
present disclosure may be based on liquid crystal technology. This
technology holds potential for providing unparallel image fidelity,
as well as increased spectral resolution in comparison to the
forensic imaging technology of the prior art. This improved
spectral resolution provides the basis for increased discrimination
and contrast enhancement capabilities. This imaging filter may be
controlled through software to transmit discrete wavelengths of
light, eliminating the need for numerous barrier or interference
filters.
[0044] In one embodiment, the system of the present disclosure may
comprise multi-conjugate filter technology available from ChemImage
Corporation, Pittsburgh Pa. This technology is more fully described
in U.S. Pat. No. 6,992,809, filed on Feb. 2, 2005, entitled
"Multi-Conjugate Liquid Crystal Tunable Filter" and U.S. Pat. No.
7,362,489, filed on Apr. 22, 2005, also entitled "Multi-Conjugate
Liquid Crystal Tunable Filter." These patents are hereby
incorporated by reference in their entireties.
[0045] The system provided for herein may utilize reflected,
transmitted, emitted, or scattered light from each point in the
sample image to create chemical-based contrast within the image.
Images are collected as a function of wavelength; therefore each
pixel within the image has a fully resolved spectrum associated
with it. By evaluating both spatial and spectral information, areas
of interest and background substrates can be resolved.
[0046] Hyperspectral imaging technology provides high spatial and
spectral resolution images for increase sensitivity and contrast
enhancement for ink discrimination. The nondestructive analysis
provides for the preservation of evidence, allowing for additional
analysis. The system requires no sample preparation, so documents
can be placed directly under the imaging optics for examination. A
custom software package may be implemented to provide for an easy
user interface that does not require examiners or be ink chemists
or specialists.
[0047] The present disclosure also provides for a method for
analyzing a questioned document. One embodiment of a method of the
present disclosure is illustrated in FIG. 3. The method 300
comprises illuminating a sample document in step 310 to thereby
generate a first plurality of interacted photons. In one
embodiment, this first plurality of interacted photons may be
photons selected from the group consisting of: photons reflected by
said sample document; photons absorbed by said sample document;
photons scattered by said sample document; photons emitted by said
sample document; and combinations thereof. In step 320 this first
plurality of interacted photons are directed to a first detector.
In step 330 this first plurality of interacted photons are detected
at the detector to thereby generate a digital image representative
of the sample document. In one embodiment, the digital image may
comprise a RGB image. This digital image is analyzed in step 340 to
thereby identify at least one region of interest of said sample. In
one embodiment, this region of interest may comprise an area of
said sample document suspected of comprising an alteration to said
sample document. This alteration may comprise at least one of: an
addition, and obliteration, a deletion, or other change to the
sample document. In another embodiment, this region of interest may
be an area of said sample document suspected of comprising more
than one type of ink. Comprising more than one type of ink may be
further evidence that an alteration has been made to a sample
document.
[0048] A region of interest is illuminated in step 350 to thereby
generate a second plurality of interacted photons. In one
embodiment, this illumination may be achieved using broadband white
light. In one embodiment, this second plurality of interacted
photons may be photons selected from the group consisting of:
photons reflected by a region of interest; photons absorbed by a
region of interest; photons scattered by a region of interest;
photons emitted by a region of interest; and combinations thereof.
In step 360 the second plurality of interacted photons are passed
through a tunable filter to thereby generate a plurality of
filtered photons. In one embodiment, these filtered photons may
comprise wavelength-selective filtered photons. The plurality of
filtered photons are directed to a second detector in step 370. In
one embodiment, this second detector may comprise a spectral
detector. These filtered photons are detected at said second
detector to thereby generate a hyperspectral image representative
of the region of interest in step 380. In one embodiment, the
hyperspectral image comprises an image and a fully resolved
spectrum unique to the material for each pixel location in said
image. In one embodiment, this hyperspectral image may comprise a
visible hyperspectral image.
[0049] In one embodiment, the hyperspectral image may be analyzed
to thereby determine if the region of interest comprises an
alteration. In another embodiment, the hyperspectral image may also
be analyzed to determine whether or not the sample comprises more
than one ink type. In another embodiment, the method may further
comprise applying one or more chemometric techniques to said
hyperspectral image. This technique may be any known in the art,
including but not limited to: principle component analysis, partial
least squares discriminate analysis, cosine correlation analysis,
Euclidian distance analysis, k-means clustering, multivariate curve
resolution, band t, entropy method, mahalanobis distance, adaptive
subspace detector, spectral mixture resolution, and combinations
thereof. In another embodiment, pattern recognition algorithms may
be used.
[0050] The present disclosure contemplates that more than one
region of interest may be identified in said sample document.
Therefore, in one embodiment, the present disclosure provides for
obtaining a hyperspectral image for each of the regions of interest
that may be present in the sample document. These hyperspectral
images can then be analyzed to determine if the region of interest
comprises an alteration. The hyperspectral image may also be
analyzed to determine of the region of interest comprises more than
one ink type.
[0051] In one embodiment of the present disclosure, the method
provided for herein may be automated or semi-automated. In one
embodiment, this automation may be achieved using software. In
another embodiment, analysis is performed on only those pixels in a
hyperspectral image comprising ink. In such an embodiment, the
background pixels (those pixels not comprising ink) are separated
from the ink pixels. This may be performed automatically or
semi-automatically. By excluding these background pixels, confusion
can be reduced, hold potential for a more discriminative result.
Therefore, in one embodiment of the present disclosure one or more
chemometric techniques may be applied to only the ink pixels.
[0052] In one embodiment of the present disclosure, preprocessing
steps may be performed in conjunction with the various embodiments
of the method disclosed herein. These preprocessing steps may
comprise prescreening a sample document and zooming in on a region
of interest. The lighting may be adjusted, acquisition conditions
defined, and sample ink data acquired. If spectralon data is
pre-acquired, then spectralon division may be done in memory when a
sample data frame is acquired. An algorithm may be utilized to
automatically detect sample pixels (pixels comprising ink) and
background pixels. An algorithm may also be used to select a subset
of background pixels and compute the average spectrum from selected
pixel spectra. This average spectrum may be used for background
division. If necessary, data may also be converted to absorbance
data.
[0053] In one embodiment of the present disclosure, the number of
analysis steps performed may rely on the data under analysis. In
one embodiment, whether or not a first, second, etc. technique is
applied may be determined by a user based on visual inspection of
the data. For example, if a sample comprises two inks with very
different characteristics, they be distinguishable based on visual
inspection of a digital image. In harder cases, it may be necessary
to perform additional analysis of hyperpsectral images by applying
processing techniques and/or chemometric techniques. For example,
k-means clustering may be applied to the data in an attempt to
enhance distinctions between inks. In another embodiment, the
results of the chemometric techniques applied may be verified by
evaluation of at least one spectra associated with at least one ink
present in the region of interest. Therefore, the present
disclosure contemplates the use of visual inspection and/or
spectral verification (which may be automatically performed) for
determination of how may chemometric techniques need to be
applied.
[0054] In harder cases, a second technique may be applied if the
first technique was not successful in clearly articulating between
ink types. In one embodiment, principal component analysis ("PCA")
may be applied to data where digital imaging was not able to
display distinctions between inks. In other harder cases, partial
least squares discriminate analysis ("PLSDA") may be applied to the
data.
[0055] In one embodiment, implementation of PLSDA may comprise
building a training set. In such an embodiment, a user may select
one or more regions of interest. In one embodiment, these regions
of interest are such that only sample pixels (pixels comprising
ink) are included. In one embodiment, the number of regions of
interest may be equal to the number of inks in a parent document.
This parent document may comprise a document wherein the number and
location of ink types are known. In one embodiment, for each region
of interest, if the number of ink pixel spectral is .gtoreq.200,
100 spectra may be randomly selected to form a class. In such an
embodiment, if the number of ink pixel is <200 then
approximately 50% of the spectra may be randomly selected. In one
embodiment, 200 background pixels (pixels not comprising ink) may
be automatically selected to form a background class.
[0056] A PLSDA model may be built using class spectra and class
number. In one embodiment, PLSDA may be applied as a two class or a
three class problem. The model may then be applied to a
hyperspectral image representative of a region of interest of the
sample document. The classification image (also referred to herein
as the "second processed image") may be displayed. A user may
evaluate the classification image to thereby determine of two or
more inks can be distinguished in the region of interest. This
evaluation may be based on visual inspection.
[0057] In one embodiment, these different analysis methods may be
used to classify inks into different groups based on the type of
ink. In one embodiment, each ink in a class may be assigned a
color. In one embodiment, this may generate what may be referred to
as a "colorized classified image."
[0058] FIG. 4 is representative of a method of the present
disclosure. The method 400 comprises identifying a region of
interest of a sample document in step 405 wherein said region of
interest comprises at least one ink type. A hyperspectral image of
the region of interest is obtained in step 410. In step 415 the
hyperspectral image is analyzed. In one embodiment, this analysis
may comprise visual inspection by a user. Visual inspection by a
user may be based on differences in color or shading between two or
more inks. If two or more ink types are distinguishable from one
another in step 415, then a result is reported in step 420a. In one
embodiment, the report in 420a may comprise a determination that
the region of interest comprises at least one of: an alteration, an
addition, a deletion, obliteration, or another change to the sample
document. The presence of two or more different inks in a sample
document may be an indication that a change has been made to the
original document.
[0059] If two or more inks are not distinguishable in step 415,
then a first chemometric technique may be applied to said
hyperspectral image to generate a first processed image
representative of said region of interest in step 420b. In one
embodiment, illustrated by FIG. 4, this first chemometric technique
may comprise k-means clustering. The first processed image is
analyzed in step 425. In one embodiment, analysis of a first
processed image may comprise visual inspection by a user. If two or
more inks are distinguishable in step 425 then a result may be
reported in step 430a. If two or more inks are not distinguishable
in step 425, then a second chemometric technique may be applied to
generate a second processed image in step 430. In one embodiment,
this second chemometric technique may comprise partial least
squares discriminate analysis ("PLSDA"). The second processed image
may be analyzed in step 435 to thereby determine if two or more
inks can be distinguished in the region of interest. If two or more
inks are distinguishable, then a result may be reported in step
440a. If two or more inks cannot be distinguished in step 435, then
an indistinguishable result may be reported in step 440b. In one
embodiment an indistinguishable result may be indicative of the
region of interest comprising only one ink type. This may infer
that either no change was made to the document, or any change made
to the document was done with the exact ink type of the original
document.
[0060] FIG. 5 is representative of a method of the present
disclosure. The method 500 comprises identifying a region of
interest of a sample document in step 505 wherein said region of
interest comprises at least one ink type. A hyperspectral image of
said region of interest is obtained in step 510. The hyperspectral
image is analyzed in step 515 to thereby determine if two or more
ink types are distinguishable in said region of interest. If two or
more ink types are distinguishable in step 515, then a result can
be reported in step 520a. If two or more ink types cannot be
distinguished in step 515, then a first chemometric technique may
be applied to the hyperspectral image to generate a first processed
image representative of the region of interest in step 520b. In one
embodiment, illustrated in FIG. 5, the first chemometric technique
may comprise k-means clustering. The first processed image may be
analyzed in step 525 to determine if two or more inks are
distinguishable in the region of interest. If two or more inks are
distinguishable, then a result may be reported in step 530a. If two
or more inks cannot be distinguished in step 525, then a second
chemometric technique may be applied to thereby generate a second
processed image in step 530b. In one embodiment, illustrated by
FIG. 5, the second chemometric technique may comprise principle
component analysis. The second processed image may be analyzed in
step 535 to thereby determine if two or more ink types are present
in said region of interest. If two or more ink types can be
distinguished in step 535, then a result may be reported in step
540a. If two or more ink types cannot be distinguished in step 535,
then an indistinguishable result may be reported in step 540b.
[0061] The present disclosure may be embodied in other specific
forms without departing from the spirit or essential attributes of
the disclosure. Accordingly, reference should be made to the
appended claims, rather than the foregoing specification, as
indicating the scope of the disclosure. Although the foregoing
description is directed to the embodiments of the disclosure, it is
noted that other variations and modifications will be apparent to
those skilled in the art, and may be made without departing from
the spirit of the disclosure.
EXAMPLE
[0062] Pen samples were prepared at ChemImage Corporation,
Pittsburgh, Pa. Five brands of pens were used: (1) Bic.RTM..; (2)
PaperMate.RTM..; (3) Pilot.RTM. (4) Universal.RTM..; and (5)
Preventa.RTM.. Each brand had several types of pens. Samples
included writing with different pens (including different pens of
the same brand and different pens of different brands), writing
with the same pens, horizontal and vertical orientations, and
negative controls. The samples were analyzed using CONDOR imaging
technology available from Chemlmage Corporation, Pittsburgh,
Pa.
[0063] K-means clustering was used to analyze all 70 data sets. For
one data set, PLSDA took 45-60 minutes to complete. Therefore, 17
data sets were selected to include different types of data and
analyzed using PLSDA. For most data sets a wavelength range of
400-720 nm, 10 nm step (33 frames) was used. For other data sets, a
wavelength range of 650-750 nm, 1 nm step (101 frames) was
used.
[0064] FIGS. 6A and 6B illustrate exemplary results for a sample
created using two different pens (pen 38 and pen 28). FIG. 6A
illustrates results after applying k-means clustering (three
classes). Visual inspection of FIG. 6A may indicate to a user that
two or more inks are present in the sample based on the changes to
the "8" in "2008". The distinctions are made even more clear in
FIG. 6B, which illustrates the result of applying PLSDA (three
classes). The three classes used were pen 38, pen 28, and the
background (no ink). The material in red corresponds to pen 38 and
the material in green corresponds to pen 28. Therefore, it is
evident from FIG. 6B that changes have been made to the sample
regarding the year, changing "2010" to "2018".
[0065] FIGS. 7A and 7B illustrate exemplary results for a sample
created using the same pen (pen 13). FIG. 7A illustrates results
after applying k-means clustering (three classes). As can be seen
in FIG. 7A, two or more inks cannot be distinguished from each
other. FIG. 7B illustrates results after applying PLSDA (three
classes). The three classes were pen 13, pen 13, and the background
(no ink). As can be seen from FIG. 7B, two or more inks cannot be
distinguished from each other.
[0066] FIGS. 8A, 8B, and 8C illustrate exemplary results for a
sample created using different pens. FIG. 8A illustrates results
after applying k-means clustering (three classes). FIG. 8B
illustrates results after applying PLSDA (three classes). The three
classes were pen 30, pen 5, and the background. FIG. 8B also
illustrates two regions of interest corresponding to two classes
(two different inks), C1 and C2. FIG. 8C illustrates the same data
as FIG. 8B but in a different orientation. As can be seen from
FIGS. 8B and 8C, PLSDA improves the ability to discriminate between
ink types. The material in green corresponds to pen 30 and the
material in red corresponds to pen 5. Therefore, it is evident that
an alteration has been made to the document, changing "31" to
"34".
* * * * *